Recently, IME (INPUT METHOD EDITOR) to input Japanese to a computer and a machine translation system are widely used. This kind of software uses a dictionary including various language information. However, content of accomplished dictionary information is predetermined. This dictionary does not include the latest word orderly generated and vocabulary used in local community such as a company. In order to solve this problem, user registers his desired word to a user dictionary. However, registration to the user dictionary by his hand is troublesome for the user. In order to simplify this registration, following two methods are used.
(1) The dictionary item is automatically registered by the user's key input or selection information when the user inputs a document.
(2) The dictionary data is automatically extracted by analyzing the accomplished document when the user indicates the accomplished document.
In the method (1), a series of key input by the user must be recognized as an operation to input a new word not previously registered in the accomplished dictionary. This operation for the series of the key input includes many kinds of variation by the user. Therefore, in order to avoid an erroneous registration, a new word should be learned from the limited pattern of key input that is certain to reflect the intention of the new word input. As a result, the coverages of the automatic registration is restricted.
In the method (2), the document is not always correctly analized and the dictionary data as the new word is not correctly extracted from the document. Especially, in IME of Japanese input, a reading as the input key is not obtained from the document.
Accordingly, in general, the user must register his desired word by his burden to the user dictionary.